Global Optimisation of Signal Settings: Meta-heuristic algorithms for solving real-scale problems

In this paper the Global Optimisation of Signal Settings (GOSS) problem is studied and a meta-heuristic algorithm is proposed for its solution. The GOSS problem arises when the parameters of all (or some) signalised intersections of a network are jointly optimised so as to minimise the value of an objective function (such as total travel time). This problem has been widely studied elsewhere and several algorithms have been proposed, mainly based on descent methods. These algorithms require high computing times for real-scale problems and usually lead to a local optimum since the objective function is hardly ever convex. The high computing times are due to the need to perform traffic assignment to determine the objective function at any iteration. In this paper we propose a multi-start method based on a Feasible Descent Direction Algorithm (FDDA) for solving this problem. The algorithm is able to search for a local optimal solution and requires lower computing times at any iteration. The proposed algorithm is tested on a real-scale network, also under different demand levels, by adopting different assignment algorithms proposed in the literature. Initial results show that the proposed algorithms perform well and that computing times are compatible with planning purposes also for real-scale networks.